Skill 寫法圖鑑

update

Anthropic | 個人與生活 | 原文 | Apache-2.0

從專案工具同步任務、整理過期項目並補齊 memory 缺口

中文翻譯

---
name: update
description: 依你目前的活動同步任務並更新 memory。要把專案追蹤工具裡的新指派拉進 TASKS.md、整理過期或逾期的任務、補上不認識的人或專案的 memory 缺口,或做全面掃描找出藏在聊天和 email 裡的待辦時使用。
argument-hint: "[--comprehensive]"
---

# Update 指令

> 如果看到不認識的 placeholder,或想確認目前連了哪些工具,請看 [CONNECTORS.md](../../CONNECTORS.md)。

讓你的任務清單和 memory 保持最新。兩種模式:

- **預設:** 從外部工具同步任務、整理過期項目、檢查 memory 的缺口
- **`--comprehensive`:** 深度掃描聊天、email、行事曆、文件,標出漏掉的待辦並建議新增 memory

## 用法

```bash
/productivity:update
/productivity:update --comprehensive
```

## 預設模式

### 1. 載入目前狀態

讀 `TASKS.md` 和 `memory/` 目錄。如果它們不存在,建議先跑 `/productivity:start`。

### 2. 從外部來源同步任務

檢查有哪些任務來源可用:
- **專案追蹤工具**(例如 Asana、Linear、Jira)(有 MCP 的話)
- **GitHub Issues**(在 repo 裡的話):`gh issue list --assignee=@me`

如果沒有可用的來源,跳到步驟 3。

**抓取指派給使用者的任務**(open/in-progress),再跟 TASKS.md 比對:

| 外部任務 | TASKS.md 有對應嗎? | 動作 |
|---------------|-----------------|--------|
| 找到,但不在 TASKS.md | 沒有對應 | 提議新增 |
| 找到,已經在 TASKS.md | 以標題比對(模糊) | 略過 |
| 在 TASKS.md,但外部沒有 | 沒有對應 | 標記為可能已過期 |
| 在外部已完成 | 在 Active 區塊 | 提議標為完成 |

把差異呈現出來,讓使用者決定要新增或完成哪些。

### 3. 整理過期項目

檢視 TASKS.md 裡的 Active 任務,標記:
- 到期日已經過了的任務
- 在 Active 放了 30 天以上的任務
- 沒有脈絡的任務(沒有人、沒有專案)

逐一呈現給使用者整理:標為完成?重新排程?移到 Someday?

### 4. 為 Memory 缺口解碼任務

對每個任務,試著解出所有實體(人、專案、縮寫、工具、連結):

```
Task: "Send PSR to Todd re: Phoenix blockers"

Decode:
- PSR → ✓ Pipeline Status Report (in glossary)
- Todd → ✓ Todd Martinez (in people/)
- Phoenix → ? Not in memory
```

記下哪些已經完全解碼、哪些還有缺口。

### 5. 補上缺口

把不認識的詞分組呈現:
```
I found terms in your tasks I don't have context for:

1. "Phoenix" (from: "Send PSR to Todd re: Phoenix blockers")
   → What's Phoenix?

2. "Maya" (from: "sync with Maya on API design")
   → Who is Maya?
```

把答案加進對應的 memory 檔案(people/、projects/、glossary.md)。

### 6. 擷取補充資訊

任務裡常常比 memory 有更豐富的脈絡。抽出來並更新:
- 任務裡的**連結** → 加到專案/人物檔案
- **狀態變更**(「launch done」)→ 更新專案狀態,從 CLAUDE.md 降級
- **關係**(「Todd 對 Maya 提案的簽核」)→ 交叉參照人物
- **截止日** → 加到專案檔案

### 7. 回報

```
Update complete:
- Tasks: +3 from project tracker (e.g. Asana), 1 completed, 2 triaged
- Memory: 2 gaps filled, 1 project enriched
- All tasks decoded ✓
```

## 全面模式(`--comprehensive`)

包含預設模式的一切,再加上對最近活動的深度掃描。

### 額外步驟:掃描活動來源

從可用的 MCP 來源蒐集資料:
- **聊天:** 搜尋最近的訊息、讀取活躍的頻道
- **Email:** 搜尋已寄出的訊息
- **文件:** 列出最近動過的文件
- **行事曆:** 列出最近和即將到來的事件

### 額外步驟:標出漏掉的待辦

把活動和 TASKS.md 比對,找出沒有被追蹤的行動項目:

```
## Possible Missing Tasks

From your activity, these look like todos you haven't captured:

1. From chat (Jan 18):
   "I'll send the updated mockups by Friday"
   → Add to TASKS.md?

2. From meeting "Phoenix Standup" (Jan 17):
   You have a recurring meeting but no Phoenix tasks active
   → Anything needed here?

3. From email (Jan 16):
   "I'll review the API spec this week"
   → Add to TASKS.md?
```

讓使用者挑要新增哪些。

### 額外步驟:建議新的 Memory

找出不在 memory 裡的新實體:

```
## New People (not in memory)
| Name | Frequency | Context |
|------|-----------|---------|
| Maya Rodriguez | 12 mentions | design, UI reviews |
| Alex K | 8 mentions | DMs about API |

## New Projects/Topics
| Name | Frequency | Context |
|------|-----------|---------|
| Starlight | 15 mentions | planning docs, product |

## Suggested Cleanup
- **Horizon project** — No mentions in 30 days. Mark completed?
```

依信心程度分組呈現。高信心的項目提議直接新增;低信心的項目先問使用者。

## 備註

- 沒有使用者確認,絕對不要自動新增任務或 memory
- 有外部來源連結的話就保留
- 任務標題用模糊比對,可以容忍用字上的小差異
- 可以放心頻繁執行,只有在有新資訊時才會更新
- `--comprehensive` 一律互動式執行

英文原文

---
name: update
description: Sync tasks and refresh memory from your current activity. Use when pulling new assignments from your project tracker into TASKS.md, triaging stale or overdue tasks, filling memory gaps for unknown people or projects, or running a comprehensive scan to catch todos buried in chat and email.
argument-hint: "[--comprehensive]"
---

# Update Command

> If you see unfamiliar placeholders or need to check which tools are connected, see [CONNECTORS.md](../../CONNECTORS.md).

Keep your task list and memory current. Two modes:

- **Default:** Sync tasks from external tools, triage stale items, check memory for gaps
- **`--comprehensive`:** Deep scan chat, email, calendar, docs — flag missed todos and suggest new memories

## Usage

```bash
/productivity:update
/productivity:update --comprehensive
```

## Default Mode

### 1. Load Current State

Read `TASKS.md` and `memory/` directory. If they don't exist, suggest `/productivity:start` first.

### 2. Sync Tasks from External Sources

Check for available task sources:
- **Project tracker** (e.g. Asana, Linear, Jira) (if MCP available)
- **GitHub Issues** (if in a repo): `gh issue list --assignee=@me`

If no sources are available, skip to Step 3.

**Fetch tasks assigned to the user** (open/in-progress). Compare against TASKS.md:

| External task | TASKS.md match? | Action |
|---------------|-----------------|--------|
| Found, not in TASKS.md | No match | Offer to add |
| Found, already in TASKS.md | Match by title (fuzzy) | Skip |
| In TASKS.md, not in external | No match | Flag as potentially stale |
| Completed externally | In Active section | Offer to mark done |

Present diff and let user decide what to add/complete.

### 3. Triage Stale Items

Review Active tasks in TASKS.md and flag:
- Tasks with due dates in the past
- Tasks in Active for 30+ days
- Tasks with no context (no person, no project)

Present each for triage: Mark done? Reschedule? Move to Someday?

### 4. Decode Tasks for Memory Gaps

For each task, attempt to decode all entities (people, projects, acronyms, tools, links):

```
Task: "Send PSR to Todd re: Phoenix blockers"

Decode:
- PSR → ✓ Pipeline Status Report (in glossary)
- Todd → ✓ Todd Martinez (in people/)
- Phoenix → ? Not in memory
```

Track what's fully decoded vs. what has gaps.

### 5. Fill Gaps

Present unknown terms grouped:
```
I found terms in your tasks I don't have context for:

1. "Phoenix" (from: "Send PSR to Todd re: Phoenix blockers")
   → What's Phoenix?

2. "Maya" (from: "sync with Maya on API design")
   → Who is Maya?
```

Add answers to the appropriate memory files (people/, projects/, glossary.md).

### 6. Capture Enrichment

Tasks often contain richer context than memory. Extract and update:
- **Links** from tasks → add to project/people files
- **Status changes** ("launch done") → update project status, demote from CLAUDE.md
- **Relationships** ("Todd's sign-off on Maya's proposal") → cross-reference people
- **Deadlines** → add to project files

### 7. Report

```
Update complete:
- Tasks: +3 from project tracker (e.g. Asana), 1 completed, 2 triaged
- Memory: 2 gaps filled, 1 project enriched
- All tasks decoded ✓
```

## Comprehensive Mode (`--comprehensive`)

Everything in Default Mode, plus a deep scan of recent activity.

### Extra Step: Scan Activity Sources

Gather data from available MCP sources:
- **Chat:** Search recent messages, read active channels
- **Email:** Search sent messages
- **Documents:** List recently touched docs
- **Calendar:** List recent + upcoming events

### Extra Step: Flag Missed Todos

Compare activity against TASKS.md. Surface action items that aren't tracked:

```
## Possible Missing Tasks

From your activity, these look like todos you haven't captured:

1. From chat (Jan 18):
   "I'll send the updated mockups by Friday"
   → Add to TASKS.md?

2. From meeting "Phoenix Standup" (Jan 17):
   You have a recurring meeting but no Phoenix tasks active
   → Anything needed here?

3. From email (Jan 16):
   "I'll review the API spec this week"
   → Add to TASKS.md?
```

Let user pick which to add.

### Extra Step: Suggest New Memories

Surface new entities not in memory:

```
## New People (not in memory)
| Name | Frequency | Context |
|------|-----------|---------|
| Maya Rodriguez | 12 mentions | design, UI reviews |
| Alex K | 8 mentions | DMs about API |

## New Projects/Topics
| Name | Frequency | Context |
|------|-----------|---------|
| Starlight | 15 mentions | planning docs, product |

## Suggested Cleanup
- **Horizon project** — No mentions in 30 days. Mark completed?
```

Present grouped by confidence. High-confidence items offered to add directly; low-confidence items asked about.

## Notes

- Never auto-add tasks or memories without user confirmation
- External source links are preserved when available
- Fuzzy matching on task titles handles minor wording differences
- Safe to run frequently — only updates when there's new info
- `--comprehensive` always runs interactively